Data Scientist & AI Engineer from Campo Grande, Brazil, with ~2 years building ML and data products end-to-end — from an LLM fiscal-audit agent over 180M+ records to fraud scoring and BI that moved real numbers. I ship the demo, not just the notebook. BSc Information Systems @ UFMS · 🇧🇷 PT (native) · 🇬🇧 EN.
| Project | Highlight | Stack |
|---|---|---|
| Julgador | Gov tax-fraud LLM audit agent — 180M+ records, 84% | Gemma 2 27B · RAG · TF |
| Perfil_Laranja | Fraud asset-risk score with no banking data — 82% | HDBSCAN · Random Forest |
| multi-agent-system (TCC) | Async orchestrator–worker AI that diagnoses a site's SEO/ads | Python · asyncio · RAG |
| face_recognition_organizer | 100% offline — ~8,886 faces → ~61 people | InsightFace · clustering |
| Demos_Repo | 20+ runnable DS/ML/NLP demos, each with metrics | scikit-learn · numpy |
→ everything else on the portfolio and all repositories.